Health R coding club

What made a hit record in 2017 according to spotify?

Hint: First things first, make sure you are working in a script rather than the console, to open a new script you can use the keyboard shortcut “Ctrl+ Shift +”.

Setting up our workspace

First we must set up R so that we are looking at the folder where we saved our data using the set working directory function: setwd(“…”)

Hint: Remember to flip your slashs, they should be /. This makes them r readable.

We must also we have all the packages we require installed.

install.packages("radarchart")
install.packages("ggplot2")
install.packages("kableExtra")
install.packages("tidyr")
install.packages("plotly")
install.packages("dplyr")
installed.packages("corrplot")

Reading in the data

 songs <- read.csv("spotify2017featuresdf.csv", header=T, sep=",")

Check your data

Let’s take a quick look at the data make sure it is as we expect.  There are a couple of key functions we can make use of to do this, these are summary and strusture.

str(songs)

## 'data.frame':    100 obs. of  16 variables:
##  $ id              : Factor w/ 100 levels "00lNx0OcTJrS3MKHcB80H",..: 97 62 53 84 42 93 46 41 7 49 ...
##  $ name            : Factor w/ 100 levels "1-800-273-8255",..: 73 21 22 80 40 39 47 94 89 41 ...
##  $ artists         : Factor w/ 78 levels "21 Savage","AJR",..: 24 48 48 71 20 40 42 26 8 76 ...
##  $ danceability    : num  0.825 0.694 0.66 0.617 0.609 0.904 0.64 0.726 0.853 0.735 ...
##  $ energy          : num  0.652 0.815 0.786 0.635 0.668 0.611 0.533 0.769 0.56 0.451 ...
##  $ key             : num  1 2 2 11 7 1 0 6 1 0 ...
##  $ loudness        : num  -3.18 -4.33 -4.76 -6.77 -4.28 ...
##  $ mode            : num  0 1 1 0 1 0 1 1 1 1 ...
##  $ speechiness     : num  0.0802 0.12 0.17 0.0317 0.0367 0.0888 0.0706 0.123 0.0406 0.0585 ...
##  $ acousticness    : num  0.581 0.229 0.209 0.0498 0.0552 0.000259 0.119 0.0293 0.013 0.0631 ...
##  $ instrumentalness: num  0.00 0.00 0.00 1.44e-05 0.00 2.03e-05 0.00 1.01e-02 0.00 1.30e-05 ...
##  $ liveness        : num  0.0931 0.0924 0.112 0.164 0.167 0.0976 0.0864 0.104 0.0944 0.325 ...
##  $ valence         : num  0.931 0.813 0.846 0.446 0.811 0.4 0.515 0.733 0.86 0.0862 ...
##  $ tempo           : num  96 88.9 177.8 103 80.9 ...
##  $ duration_ms     : num  233713 228827 228200 247160 288600 ...
##  $ time_signature  : num  4 4 4 4 4 4 4 4 4 4 ...

This function tells us about the structure of our data, here we can see the different variables within our data and the the variable type for example our data has numeric and factor variables.

summary(songs)

##                      id                           name   
##  00lNx0OcTJrS3MKHcB80H: 1   1-800-273-8255          : 1  
##  04DwTuZ2VBdJCCC5TROn7: 1   24K Magic               : 1  
##  0afhq8XCExXpqazXczTSv: 1   2U (feat. Justin Bieber): 1  
##  0CcQNd8CINkwQfe1RDtGV: 1   Ahora Dice              : 1  
##  0CokSRCu5hZgPxcZBaEzV: 1   All Night               : 1  
##  0dA2Mk56wEzDgegdC6R17: 1   Attention               : 1  
##  (Other)              :94   (Other)                 :94  
##              artists    danceability        energy            key       
##  Ed Sheeran      : 4   Min.   :0.2580   Min.   :0.3460   Min.   : 0.00  
##  The Chainsmokers: 4   1st Qu.:0.6350   1st Qu.:0.5565   1st Qu.: 2.00  
##  Drake           : 3   Median :0.7140   Median :0.6675   Median : 6.00  
##  Martin Garrix   : 3   Mean   :0.6968   Mean   :0.6607   Mean   : 5.57  
##  Bruno Mars      : 2   3rd Qu.:0.7702   3rd Qu.:0.7875   3rd Qu.: 9.00  
##  Calvin Harris   : 2   Max.   :0.9270   Max.   :0.9320   Max.   :11.00  
##  (Other)         :82                                                    
##     loudness            mode       speechiness       acousticness     
##  Min.   :-11.462   Min.   :0.00   Min.   :0.02320   Min.   :0.000259  
##  1st Qu.: -6.595   1st Qu.:0.00   1st Qu.:0.04312   1st Qu.:0.039100  
##  Median : -5.437   Median :1.00   Median :0.06265   Median :0.106500  
##  Mean   : -5.653   Mean   :0.58   Mean   :0.10397   Mean   :0.166306  
##  3rd Qu.: -4.327   3rd Qu.:1.00   3rd Qu.:0.12300   3rd Qu.:0.231250  
##  Max.   : -2.396   Max.   :1.00   Max.   :0.43100   Max.   :0.695000  
##                                                                       
##  instrumentalness       liveness          valence           tempo       
##  Min.   :0.000e+00   Min.   :0.04240   Min.   :0.0862   Min.   : 75.02  
##  1st Qu.:0.000e+00   1st Qu.:0.09828   1st Qu.:0.3755   1st Qu.: 99.91  
##  Median :0.000e+00   Median :0.12500   Median :0.5025   Median :112.47  
##  Mean   :4.796e-03   Mean   :0.15061   Mean   :0.5170   Mean   :119.20  
##  3rd Qu.:1.335e-05   3rd Qu.:0.17925   3rd Qu.:0.6790   3rd Qu.:137.17  
##  Max.   :2.100e-01   Max.   :0.44000   Max.   :0.9660   Max.   :199.86  
##                                                                         
##   duration_ms     time_signature
##  Min.   :165387   Min.   :3.00  
##  1st Qu.:198491   1st Qu.:4.00  
##  Median :214106   Median :4.00  
##  Mean   :218387   Mean   :3.99  
##  3rd Qu.:230543   3rd Qu.:4.00  
##  Max.   :343150   Max.   :4.00  
## 

Here we can see the ranges of numerical values. 

We can also take a look at our full data set:

library(kableExtra)
kable(songs) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed", "responsive"), font_size = 12) %>%
scroll_box(width = "720px", height = "500px")

To enable us to know how popular a song is i.e. where t placed in the top 100, we must add a variable to indicate a songs position.

songs$rank <- seq.int(nrow(songs))

Feature exploration

teal <- "#17bebb"
speech <- ggplot(songs, aes(x=reorder(name, -speechiness), y=speechiness, text=(paste( "Track:", name, "<br>", "Artist:", artists, "<br>", "Speechiness:", speechiness))))+
  geom_col(fill=teal)+
  theme_minimal()+
  theme(axis.title.x=element_blank(),
        axis.text.x=element_blank(),
        axis.ticks.x=element_blank(),
        axis.ticks.y=element_blank())+
  ggtitle("Speechiness")
ggplotly(speech, tooltip=c("text"))

pink <- "#ff6f59"
dance <- ggplot(songs, aes(x=reorder(name, -danceability), y=danceability, text=(paste( "Track:", name, "<br>", "Artist:", artists, "<br>", "Danceability:", danceability))))+
  geom_col(fill=pink)+
  theme_minimal()+
  theme(axis.title.x=element_blank(),
        axis.text.x=element_blank(),
        axis.ticks.x=element_blank(),
        axis.ticks.y=element_blank())+
  ggtitle("Danceability")
ggplotly(dance, tooltip=c("text"))

songs_key <- songs%>%
  select(name, key)%>%
  mutate(n=n())%>%
  unique()%>%
  group_by(key)%>%
  mutate(total=sum(n))%>%
  mutate(percent=round((n/total)*100))

head(songs_key, 10)

## # A tibble: 10 x 5
## # Groups:   key [6]
##    name                                            key     n total percent
##    <fct>                                         <dbl> <int> <int>   <dbl>
##  1 Shape of You                                      1   100  1400       7
##  2 Despacito - Remix                                 2   100   800      12
##  3 Despacito (Featuring Daddy Yankee)                2   100   800      12
##  4 Something Just Like This                         11   100  1300       8
##  5 I'm the One                                       7   100   700      14
##  6 HUMBLE.                                           1   100  1400       7
##  7 It Ain't Me (with Selena Gomez)                   0   100  1000      10
##  8 Unforgettable                                     6   100   900      11
##  9 That's What I Like                                1   100  1400       7
## 10 "I Don’t Wanna Live Forever (Fifty Shades ~     0   100  1000      10

key <- ggplot(songs_key, aes(x=key))

songs_5 <- songs[1:5,-c(1,3,17)]
songs_5_n <- cbind(songs_5[1], apply(songs_5[-1],2,function(x){(x-min(x))/diff(range(x))}))
radarDF <- gather(songs_5_n, key=Attribute, value=Score, -name) %>%
  spread(key=name, value=Score)

chartJSRadar(scores = radarDF,
             scaleStartValue = -1,
             maxScale=1,
             showToolTipLabel=TRUE)

songs_5 <- songs[1:5,-c(1,3,17,16)]
songs_5_n <- cbind(songs_5[1], apply(songs_5[-1],2,function(x){(x-min(x))/diff(range(x))}))
radarDF <- gather(songs_5_n, key=Attribute, value=Score, -name) %>%
  spread(key=name, value=Score)

chartJSRadar(scores = radarDF,
             scaleStartValue = -1,
             maxScale=1,
             showToolTipLabel=TRUE)